Chai Discovery has announced a collaboration agreement under which argenx will gain early access to Chai’s AI platform for designing antibodies from scratch. Chai published the announcement on July 15, 2026. The public details are thin: the release names no financial terms, therapeutic targets or results, so it shows intent, not outcomes.
Table of Contents
What Chai Discovery and argenx announced
In its release, Chai describes a “collaboration agreement” with argenx, which it calls a leading immunology innovation company. The subhead says argenx “will gain early access” to Chai’s AI platform to support de novo antibody discovery against therapeutic targets. The Business Wire version carries the same text.
The wording is forward-looking. It describes access argenx will get, not work already under way. Nothing publicly available confirms that argenx scientists have started using the platform, or that any antibody has come out of the collaboration.
The official headline is “Chai Discovery Enters Into Collaboration Agreement With argenx to Advance AI-Driven Immunology Breakthroughs.”
What each company said
Both quotes in the release are general.
Joshua Meier, Chai’s co-founder and CEO, said argenx is the kind of scientifically rigorous, fast-moving team that can make real use of frontier AI models. He added that Chai sees an opportunity to give argenx scientists de novo design capabilities for key targets of interest.
Peter Ulrichts, argenx’s chief scientific officer, said the company’s innovation model combines disease biology, antibody engineering and rigorous experimentation. He called AI-driven molecular design a powerful extension of that approach.
Neither statement says which targets are involved, what will be designed or when. Next AI Press did not find a separate announcement from argenx. The company’s newsroom index showed no Chai release, though its full press-release list did not load, so that absence is not conclusive.
The table below separates what comes from Chai from what comes from elsewhere.
| Statement | Source | Type |
|---|---|---|
| argenx will gain early access to Chai’s AI platform | Chai release | Company statement |
| Chai-2 was the first zero-shot antibody design platform with double-digit hit rates, a 100-fold improvement | Chai release | Company claim |
| Chai-3 further improves performance | Chai release | Company claim; no benchmarks verified |
| 16% hit rate in fully de novo design, given a defined epitope | Fierce Biotech, reporting a Chai preprint | Outlet reporting a company preprint; not peer-reviewed |
| No financial terms, targets or indications named | Chai release; AllSci | Company release; secondary summary |
| Evidence that argenx is using the platform, or of results | None found | n/a |
How de novo antibody design works
Antibody drugs are traditionally found by screening large libraries of candidates or by modifying existing antibodies. “De novo” design instead aims to generate new antibodies computationally, without a pre-existing template, and then test the AI-proposed designs in the lab.
Chai says its models are built to move discovery teams “beyond screening and optimization” toward high-precision molecular design. Its release describes its earlier model, Chai-2, released in 2025, as the first zero-shot antibody design platform to reach double-digit experimental hit rates. Chai calls that a 100-fold improvement over earlier computational approaches. Those are Chai’s own characterizations.
Some independent context comes from Fierce Biotech, which reported that Chai backed its claims in preprint papers. One found a 16% hit rate in fully de novo antibody design when given a defined epitope on a target. Two caveats apply. Preprints have not been through peer review. And the result comes from Chai’s own preprint, so it does not show how the models will perform on argenx’s targets.
Which Chai model does argenx get?
That is not clear. The release mentions Chai-3, which it describes as Chai’s latest model and says improves on several fronts, including hard-to-drug targets and multispecific design. But it says argenx will get access to “Chai’s AI platform,” and it does not say Chai-3 is included. By contrast, Chai’s news page states explicitly that Pfizer’s scientists will use Chai-3.
The Chai-3 improvements are also Chai’s own claim. Public benchmark figures for Chai-3 were not located, so none are cited here.
What has not been disclosed
The release leaves out most of the details that would show how substantial the arrangement is.
| Item | What is public |
|---|---|
| Agreement type | Collaboration agreement (Chai release) |
| Access | argenx “will gain early access” to Chai’s AI platform (Chai release) |
| Purpose | De novo antibody discovery “against therapeutic targets” (Chai release) |
| Financial terms and milestones | Not publicly disclosed |
| Named targets, programs or indications | Not publicly disclosed |
| Model version (Chai-2, Chai-3 or other) | Not specified for argenx |
| Duration, exclusivity, IP ownership | Not publicly disclosed |
| Deployment model (hosted, custom, other) | Not publicly disclosed |
| Results | None reported |
A secondary outlet, AllSci, also reports that no financial terms, molecules, targets or indications were named. It is a summary, not independent reporting, and it says it was produced with AI assistance.
“Not disclosed” does not mean the terms don’t exist. It means the public cannot see them.
Where this fits in Chai’s pharma roster
Chai has been announcing agreements with large drugmakers at a fast pace. Fierce Biotech’s coverage of Chai’s $400M Series C, which values the company at $3.8 billion, reports that Chai struck deals with Eli Lilly, Pfizer and Novartis. Chai later announced a collaboration with Bristol Myers Squibb, according to a Business Wire release.
The argenx agreement is one of those. It should not be read through the details of the others. Some of the other agreements have been described as involving custom models or specific model access, and nothing verified says the argenx agreement includes those elements.
Is this validation?
Chai’s release presents the collaboration as part of a broader shift in which frontier AI models move from research into use in real workflows. That is Chai’s framing.
The public evidence supports a narrower reading. An immunology company that Chai describes as a leader in its field has agreed to a collaboration involving Chai’s platform. That signals interest. It does not show that the platform performs well on argenx’s targets, and no results have been disclosed. Whether the tools will produce antibodies argenx wants to advance is the open question, and it will take work and time to answer.
Latest developments and what to watch
As of September 19, 2026, no update from argenx about the collaboration was found, and no results, expansion or change have been disclosed. Chai’s own later news concerns other partners.
Points worth watching:
- Whether argenx says anything about the collaboration in its own communications or filings.
- Whether either company discloses targets, programs or model access.
- Whether any results appear, ideally with peer-reviewed data.
FAQ
What did Chai Discovery announce with argenx? A collaboration agreement, announced July 15, 2026, under which argenx will gain early access to Chai’s AI platform for de novo antibody discovery.
What will argenx get access to? Chai’s release says “Chai’s AI platform.” It does not specify the model, the deployment approach or the scope of use.
Were financial terms or targets disclosed? No. The release names neither, and a secondary report says the same.
Is argenx already using Chai’s platform? That is unconfirmed. The release says argenx “will gain” early access. No public evidence shows usage or results.
Does the deal include Chai-3? Not stated. The release mentions Chai-3 in describing the company but does not say argenx will use it.
What is de novo antibody discovery, and what hit rate does Chai claim for Chai-2? It is computational design of new antibodies without a template. Chai says Chai-2 was first to reach double-digit experimental hit rates. Fierce reports a preprint result of 16% given a defined epitope. This is a company-reported figure from a non-peer-reviewed preprint.
Has argenx made its own announcement? None was found, though argenx’s full press-release list did not load, so this is not conclusive.
Conclusion
What is confirmed is limited: Chai Discovery announced a collaboration agreement, and argenx will gain early access to its AI platform for de novo antibody design. Most of what would show how much it matters is not public: the terms, the targets, the model argenx will use and any results. Chai’s performance claims for Chai-2 and Chai-3 are its own, and the 16% hit rate Fierce Biotech reported comes from a preprint, not from work with argenx.
The next signals to watch are any disclosure from argenx, any named targets or programs, and any data from work done under the agreement. Until then, the announcement is best read as an early step.